arXiv subjects

Li Li

Publications and source records attributed to Li Li.

At least 163 records · Page 9Linked to original sources

arXiv2025

Tree of Agents: Improving Long-Context Capabilities of Large Language Models through Multi-Perspective Reasoning

Large language models (LLMs) face persistent challenges when handling long-context tasks, most notably the lost in the middle issue, where information located in the middle of a long input tends to be underutilized. Some existing methods that reduce input have the risk of discarding key information, while others that extend context windows often lead to attention dispersion. To address these limitations, we propose Tree of Agents (TOA), a multi-agent reasoning framework that segments the input into chunks processed by independent agents. Each agent generates its local cognition, then agents dynamically exchange information for collaborative reasoning along tree-structured paths. TOA enables agents to probe different reasoning orders for multi-perspective understanding, effectively mitigating position bias and reducing hallucinations. To improve processing efficiency, we incorporate prefix-hash caching and adaptive pruning strategies, achieving significant performance improvements with comparable API overhead. Experiments show that TOA, powered by compact LLaMA3.1-8B, significantly outperforms multiple baselines and demonstrates comparable performance to the latest and much larger commercial models, such as Gemini1.5-pro, on various long-context tasks. Code is available at https://github.com/Aireduce952/Tree-of-Agents.

cs.AI↗

arXiv2025

Holographic study of shear viscosity and butterfly velocity for magnetic field-driven quantum criticality

We investigate the shear viscosity and butterfly velocity of a magnetic field-induced quantum phase transition in five dimensional Einstein-Maxwell-Chern-Simons theory, which is holographically dual to a class of strongly coupled quantum field theories with chiral anomalies. Our analysis reveals that the ratio of longitudinal shear viscosity to entropy density $η_\parallel/s$ exhibits a pronounced non-monotonic dependence on temperature $T$ when the magnetic field $B$ is slightly below the critical value $B_c$ of the quantum phase transition. In particular, it can develop a distinct minimum at an intermediate temperature. This contrasts sharply with the monotonic temperature scaling observed at and above $B_c$, where $η_\parallel/s$ follows the scaling $T^{2/3}$ at $B=B_c$ and transitions to $T^2$ for $B>B_c$ as $T\to0$. The non-vanishing of $η_\parallel/s$ for $B<B_c$ in the zero temperature limit suggests that it could serve as a good order parameter of the quantum phase transition. We also find that all butterfly velocities change dramatically near the quantum phase transition, and thus their derivatives with respect to $B$ can be independently used to detect the quantum critical point.

hep-th↗

arXiv2025

Enhance Large Language Models as Recommendation Systems with Collaborative Filtering

As powerful tools in Natural Language Processing (NLP), Large Language Models (LLMs) have been leveraged for crafting recommendations to achieve precise alignment with user preferences and elevate the quality of the recommendations. The existing approaches implement both non-tuning and tuning strategies. Compared to following the tuning strategy, the approaches following the non-tuning strategy avoid the relatively costly, time-consuming, and expertise-requiring process of further training pre-trained LLMs on task-specific datasets, but they suffer the issue of not having the task-specific business or local enterprise knowledge. To the best of our knowledge, none of the existing approaches following the non-tuning strategy explicitly integrates collaborative filtering, one of the most successful recommendation techniques. This study aims to fill the gap by proposing critique-based LLMs as recommendation systems (Critic-LLM-RS). For our purpose, we train a separate machine-learning model called Critic that implements collaborative filtering for recommendations by learning from the interactions between many users and items. The Critic provides critiques to LLMs to significantly refine the recommendations. Extensive experiments have verified the effectiveness of Critic-LLM-RS on real datasets.

cs.IR↗

arXiv2025

Bridging Memory Gaps: Scaling Federated Learning for Heterogeneous Clients

Federated Learning (FL) enables multiple clients to collaboratively train a shared model while preserving data privacy. However, the high memory demand during model training severely limits the deployment of FL on resource-constrained clients. To this end, we propose \our, a scalable and inclusive FL framework designed to overcome memory limitations through sequential block-wise training. The core idea of \our is to partition the global model into blocks and train them sequentially, thereby reducing training memory requirements. To mitigate information loss during block-wise training, \our introduces a Curriculum Mentor that crafts curriculum-aware training objectives for each block to steer their learning process. Moreover, \our incorporates a Training Harmonizer that designs a parameter co-adaptation training scheme to coordinate block updates, effectively breaking inter-block information isolation. Extensive experiments on both simulation and hardware testbeds demonstrate that \our significantly improves model performance by up to 84.2\%, reduces peak memory usage by up to 50.4\%, and accelerates training by up to 1.9$\times$.

cs.DC↗

arXiv2025

Effective Two-Stage Double Auction for Dynamic Resource Provision over Edge Networks via Discovering The Power of Overbooking

To facilitate responsive and cost-effective computing service delivery over edge networks, this paper investigates a novel two-stage double auction methodology via discovering an interesting idea of resource overbooking to overcome dynamic and uncertain nature of supply of edge servers (sellers) and demand generated from mobile devices (as buyers). The proposed auction integrates multiple essential goals such as maximizing social welfare as well as accelerating the decision-making process from both short-term and long-term views, (e.g., the time for determining winning seller-buyer pairs), by introducing a stagewise strategy: an overbooking-driven pre-double auction (OPDAuction) for determining long-term cooperations between sellers and buyers before practical resource transactions as Stage I, and a real-time backup double auction (RBDAuction) for quickly coping with residual resource demands during actual transactions. In particular, by embedding a proper overbooking rate, OPDAuction helps with facilitating trading contracts between appropriate sellers and buyers as guidance for future transactions, by allowing the booked resources to exceed theoretical supply. Then, since pre-auctions may cause risks, our RBDAuction adjusts to real-time market changes, further enhancing the overall social welfare. More importantly, we offer an interesting view to show that our proposed two-stage auction can support significant design properties such as truthfulness, individual rationality, and budget balance. Through extensive experiments, we demonstrate good performance in social welfare, time efficiency, and computational scalability, outstripping conventional methods in dynamic edge computing settings.

cs.GT↗

arXiv2025

FedHybrid: Breaking the Memory Wall of Federated Learning via Hybrid Tensor Management

Federated Learning (FL) emerges as a new learning paradigm that enables multiple devices to collaboratively train a shared model while preserving data privacy. However, one fundamental and prevailing challenge that hinders the deployment of FL on mobile devices is the memory limitation. This paper proposes \textit{FedHybrid}, a novel framework that effectively reduces the memory footprint during the training process while guaranteeing the model accuracy and the overall training progress. Specifically, \textit{FedHybrid} first selects the participating devices for each training round by jointly evaluating their memory budget, computing capability, and data diversity. After that, it judiciously analyzes the computational graph and generates an execution plan for each selected client in order to meet the corresponding memory budget while minimizing the training delay through employing a hybrid of recomputation and compression techniques according to the characteristic of each tensor. During the local training process, \textit{FedHybrid} carries out the execution plan with a well-designed activation compression technique to effectively achieve memory reduction with minimum accuracy loss. We conduct extensive experiments to evaluate \textit{FedHybrid} on both simulation and off-the-shelf mobile devices. The experiment results demonstrate that \textit{FedHybrid} achieves up to a 39.1\% increase in model accuracy and a 15.5$\times$ reduction in wall clock time under various memory budgets compared with the baselines.

cs.LG↗

arXiv2025

Probability Distribution Learning and Its Application in Deep Learning

Despite its empirical success, deep learning still lacks a comprehensive theoretical understanding of model fitting and generalization. This paper proposes the probability distribution (PD) learning framework to analyze the optimization and generalization mechanisms of deep learning. Within this framework, the conditional distribution of labels given features is the primary learning target, with the loss function, prior knowledge, and model properties explicitly characterized. Under these formulations, we establish theoretical guarantees on optimizability, even in non-convex settings, and derive generalization error bounds that provide meaningful explanations for practical performance. Specifically, we first prove theoretically that the Fenchel-Young loss is the natural and necessary choice for solving PD learning problems, thereby justifying the generality of conclusions based on this loss. Second, to capture the characteristics of deep neural networks (DNNs), we introduce the notions of $\mathcal{H}(ψ)$-convexity and $\mathcal{H}(Ψ)$-smoothness, which generalize the classical concepts of strong convexity and Lipschitz smoothness. Based on them, we provide a theoretical explanation for the effectiveness of SGD in training DNNs. Finally, we derive model-independent bounds on the expected risk and generalization error for trained models, revealing the influence of the training set size, regularization term, the mutual information between labels and features, and the information loss caused by model irreversibility on risk and generalization. Based on our theoretical analysis and experimental validation, we believe that the PD learning framework facilitates a deeper and more unified theoretical understanding of deep learning.

cs.LG↗

arXiv2025

CoT Vectors: Transferring and Probing the Reasoning Mechanisms of LLMs

Chain-of-Thought (CoT) prompting has emerged as a powerful approach to enhancing the reasoning capabilities of Large Language Models (LLMs). However, existing implementations, such as in-context learning and fine-tuning, remain costly and inefficient. To improve CoT reasoning at a lower cost, and inspired by the task vector paradigm, we introduce CoT Vectors, compact representations that encode task-general, multi-step reasoning knowledge. Through experiments with Extracted CoT Vectors, we observe pronounced layer-wise instability, manifesting as a U-shaped performance curve that reflects a systematic three-stage reasoning process in LLMs. To address this limitation, we propose Learnable CoT Vectors, optimized under a teacher-student framework to provide more stable and robust guidance. Extensive evaluations across diverse benchmarks and models demonstrate that CoT Vectors not only outperform existing baselines but also achieve performance comparable to parameter-efficient fine-tuning methods, while requiring fewer trainable parameters. Moreover, by treating CoT Vectors as a probe, we uncover how their effectiveness varies due to latent space structure, information density, acquisition mechanisms, and pre-training differences, offering new insights into the functional organization of multi-step reasoning in LLMs. The source code will be released.

cs.CL↗

arXiv2025

FedGCS: A Generative Framework for Efficient Client Selection in Federated Learning via Gradient-based Optimization

Federated Learning faces significant challenges in statistical and system heterogeneity, along with high energy consumption, necessitating efficient client selection strategies. Traditional approaches, including heuristic and learning-based methods, fall short of addressing these complexities holistically. In response, we propose FedGCS, a novel generative client selection framework that innovatively recasts the client selection process as a generative task. Drawing inspiration from the methodologies used in large language models, FedGCS efficiently encodes abundant decision-making knowledge within a continuous representation space, enabling efficient gradient-based optimization to search for optimal client selection that will be finally output via generation. The framework comprises four steps: (1) automatic collection of diverse "selection-score" pair data using classical client selection methods; (2) training an encoder-evaluator-decoder framework on this data to construct a continuous representation space; (3) employing gradient-based optimization in this space for optimal client selection; (4) generating the final optimal client selection via using beam search for the well-trained decoder. FedGCS outperforms traditional methods by being more comprehensive, generalizable, and efficient, simultaneously optimizing for model performance, latency, and energy consumption. The effectiveness of FedGCS is proven through extensive experimental analyses.

cs.LG↗

arXiv2025

Experimental Extraction of Coherent Ergotropy and Its Energetic Cost in a Superconducting Qubit

Quantum coherence, encoded in the off-diagonal elements of a system's density matrix, is a key resource in quantum thermodynamics, fundamentally limiting the maximum extractable work known as ergotropy. While previous experiments have isolated coherence-related contributions to work extraction, it remains unclear how coherence can be harnessed in a controllable and energy-efficient manner. Here, we experimentally investigate the role of initial-state coherence in work extraction from a superconducting transmon qubit. By preparing a variety of pure states and implementing three complementary extraction protocols, we reveal how coherence governs the partitioning of ergotropy. We find that the choice of initial state depends on the dominant decoherence channel-energy relaxation or dephasing. By further accounting for thermodynamic costs, we identify optimal initial states that maximize the efficiency. Our results demonstrate that the initial-state design provides a scalable approach to coherence control and advances the development of efficient quantum thermodynamic devices.

quant-ph↗

arXiv2025

Stability analysis for $Q$-balls with spectral method

Based on the spectral decomposition technique, we introduce a simple and universal numerical method to analyze the stability of solitons. Adopting this method, the linear dynamical properties of $Q$-balls are systematically revealed, from the fundamental to the excited states. For the fundamental $Q$-ball, the well-known stability criterion holds. However, for the excited $Q$-balls, the situation becomes extremely complicated, in which the stability criterion is violated. The system exhibits dynamical instability to both spherically symmetric and non-spherically symmetric perturbations, manifested in the appearance of complex and imaginary modes. In addition, we observe two interesting phenomena. One is that the oscillation mode and the complex or imaginary mode can transform into each other, marking the transition of the dynamical properties of the system. The other is the existence of excited $Q$-balls capable of resisting perturbations with low-order spherical harmonics. Such results indicate that the excited $Q$-balls will exhibit rich dynamical behaviors.

hep-th↗

arXiv2025

Towards Seamless Hierarchical Federated Learning under Intermittent Client Participation: A Stagewise Decision-Making Methodology

Federated Learning (FL) offers a pioneering distributed learning paradigm that enables devices/clients to build a shared global model. This global model is obtained through frequent model transmissions between clients and a central server, which may cause high latency, energy consumption, and congestion over backhaul links. To overcome these drawbacks, Hierarchical Federated Learning (HFL) has emerged, which organizes clients into multiple clusters and utilizes edge nodes (e.g., edge servers) for intermediate model aggregations between clients and the central server. Current research on HFL mainly focus on enhancing model accuracy, latency, and energy consumption in scenarios with a stable/fixed set of clients. However, addressing the dynamic availability of clients -- a critical aspect of real-world scenarios -- remains underexplored. This study delves into optimizing client selection and client-to-edge associations in HFL under intermittent client participation so as to minimize overall system costs (i.e., delay and energy), while achieving fast model convergence. We unveil that achieving this goal involves solving a complex NP-hard problem. To tackle this, we propose a stagewise methodology that splits the solution into two stages, referred to as Plan A and Plan B. Plan A focuses on identifying long-term clients with high chance of participation in subsequent model training rounds. Plan B serves as a backup, selecting alternative clients when long-term clients are unavailable during model training rounds. This stagewise methodology offers a fresh perspective on client selection that can enhance both HFL and conventional FL via enabling low-overhead decision-making processes. Through evaluations on MNIST and CIFAR-10 datasets, we show that our methodology outperforms existing benchmarks in terms of model accuracy and system costs.

cs.LG↗

arXiv2025

In-Loop Filtering Using Learned Look-Up Tables for Video Coding

In-loop filtering (ILF) is a key technology in video coding standards to reduce artifacts and enhance visual quality. Recently, neural network-based ILF schemes have achieved remarkable coding gains, emerging as a powerful candidate for next-generation video coding standards. However, the use of deep neural networks (DNN) brings significant computational and time complexity or high demands for dedicated hardware, making it challenging for general use. To address this limitation, we study a practical ILF solution by adopting look-up tables (LUTs). After training a DNN with a restricted reference range for ILF, all possible inputs are traversed, and the output values of the DNN are cached into LUTs. During the coding process, the filtering process is performed by simply retrieving the filtered pixel through locating the input pixels and interpolating between the cached values, instead of relying on heavy inference computations. In this paper, we propose a universal LUT-based ILF framework, termed LUT-ILF++. First, we introduce the cooperation of multiple kinds of filtering LUTs and propose a series of customized indexing mechanisms to enable better filtering reference perception with limited storage consumption. Second, we propose the cross-component indexing mechanism to enable the filtering of different color components jointly. Third, in order to make our solution practical for coding uses, we propose the LUT compaction scheme to enable the LUT pruning, achieving a lower storage cost of the entire solution. The proposed framework is implemented in the VVC reference software. Experimental results show that the proposed framework achieves on average 0.82%/2.97%/1.63% and 0.85%/4.11%/2.06% bitrate reduction for common test sequences, under the AI and RA configurations, respectively. Compared to DNN-based solutions, our proposed solution has much lower time complexity and storage cost.

eess.IV↗

arXiv2025

Thermalization dynamics of finite-size quantum critical systems

Using holographic duality, we investigate thermalization process when two finite-size quantum critical systems are brought into thermal contact along a perfectly transmitting interface. Through real-time simulations of gravitational dynamics, which are spatially inhomogeneous and anisotropic and are confined within two dynamical bulk branes, we identify three distinct thermalization patterns governed by the energy imbalance (temperature difference) and system size. For systems with large size and small energy imbalance, we observe recurrent cycles of formation and collapse of non-equilibrium steady states (NESS). Under large energy imbalance, shock waves persist for a prolonged period with sustained boundary reflections, while rarefaction waves rapidly homogenize. When the system size is sufficiently small, dissipation dominates and leads to oscillatory decay without sustained NESS or shock structure. In sharp contrast to diffusive systems, we uncover that wave-propagated energy transfer together with boundary reflections enables nearly complete energy swapping between subsystems during thermalization. Our results reveal rich thermalization dynamics in finite-size quantum critical systems across spatial scales and energy gradient regimes.

hep-th↗

arXiv2025

Recent research on $(-1)$-homogeneous solutions of stationary Navier-Stokes equations

We make an exposition of recent research on $(-1)$-homogeneous solutions of the three-dimensional incompressible stationary Navier-Stokes equations with singular rays. We also discuss properties of such solutions that are axisymmetric with no swirl, and present graphs illustrating examples that exhibit various typical types of singular behavior.

math.AP↗

arXiv2025

Dense Matter in Neutron Stars with eXTP

In this White Paper, we present the potential of the enhanced X-ray Timing and Polarimetry (eXTP) mission to constrain the equation of state of dense matter in neutron stars, exploring regimes not directly accessible to terrestrial experiments. By observing a diverse population of neutron stars - including isolated objects, X-ray bursters, and accreting systems - eXTP's unique combination of timing, spectroscopy, and polarimetry enables high-precision measurements of compactness, spin, surface temperature, polarimetric signals, and timing irregularity. These multifaceted observations, combined with advances in theoretical modeling, pave the way toward a comprehensive description of the properties and phases of dense matter from the crust to the core of neutron stars. Under development by an international Consortium led by the Institute of High Energy Physics of the Chinese Academy of Sciences, the eXTP mission is planned to be launched in early 2030.

astro-ph.HE↗

arXiv2025

UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning

The development of autonomous agents for graphical user interfaces (GUIs) presents major challenges in artificial intelligence. While recent advances in native agent models have shown promise by unifying perception, reasoning, action, and memory through end-to-end learning, open problems remain in data scalability, multi-turn reinforcement learning (RL), the limitations of GUI-only operation, and environment stability. In this technical report, we present UI-TARS-2, a native GUI-centered agent model that addresses these challenges through a systematic training methodology: a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment that integrates file systems and terminals, and a unified sandbox platform for large-scale rollouts. Empirical evaluation demonstrates that UI-TARS-2 achieves significant improvements over its predecessor UI-TARS-1.5. On GUI benchmarks, it reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld, outperforming strong baselines such as Claude and OpenAI agents. In game environments, it attains a mean normalized score of 59.8 across a 15-game suite-roughly 60% of human-level performance-and remains competitive with frontier proprietary models (e.g., OpenAI o3) on LMGame-Bench. Additionally, the model can generalize to long-horizon information-seeking tasks and software engineering benchmarks, highlighting its robustness across diverse agent tasks. Detailed analyses of training dynamics further provide insights into achieving stability and efficiency in large-scale agent RL. These results underscore UI-TARS-2's potential to advance the state of GUI agents and exhibit strong generalization to real-world interactive scenarios.

cs.AI↗

arXiv2025

EHVC: Efficient Hierarchical Reference and Quality Structure for Neural Video Coding

Neural video codecs (NVCs), leveraging the power of end-to-end learning, have demonstrated remarkable coding efficiency improvements over traditional video codecs. Recent research has begun to pay attention to the quality structures in NVCs, optimizing them by introducing explicit hierarchical designs. However, less attention has been paid to the reference structure design, which fundamentally should be aligned with the hierarchical quality structure. In addition, there is still significant room for further optimization of the hierarchical quality structure. To address these challenges in NVCs, we propose EHVC, an efficient hierarchical neural video codec featuring three key innovations: (1) a hierarchical multi-reference scheme that draws on traditional video codec design to align reference and quality structures, thereby addressing the reference-quality mismatch; (2) a lookahead strategy to utilize an encoder-side context from future frames to enhance the quality structure; (3) a layer-wise quality scale with random quality training strategy to stabilize quality structures during inference. With these improvements, EHVC achieves significantly superior performance to the state-of-the-art NVCs. Code will be released in: https://github.com/bytedance/NEVC.

eess.IV↗